US11858627B2ActiveUtilityA1
Method of high energy efficiency unmanned aerial vehicle (UAV) green data acquisition system
Est. expiryJan 29, 2041(~14.6 yrs left)· nominal 20-yr term from priority
B64U 2101/00B64C 39/024G05B 17/02B64U 2201/10H04W 4/38B64U 10/10H04W 72/0446H04W 72/0473Y02D30/70B64U 2101/20
39
PatentIndex Score
0
Cited by
7
References
1
Claims
Abstract
A design method of a high energy efficiency unmanned aerial vehicle (UAV) green data acquisition system belongs to the technical field of data acquisition and optimization for UAV uplink communication. Firstly, a system optimization objective is constructed; and in a uplink communication network of a single UAV and ground sensors, the UAV receives data periodically. Secondly, according to a constructed optimization problem, the optimization objective is maximization of EE({W},{t},{S}). Finally, an original problem is decomposed into two approximate concave-convex fractional sub-problem based on a block coordinate descent method and a successive convex approximation technique to obtain a suboptimal solution; an overall iterative algorithm is proposed: in each iteration, by solving the sub-problems, wake-up scheduling S, time slot t and UAV trajectory W are alternately optimized. The solution obtained in each iteration is used as the input of next iteration. The present invention can jointly optimize the UAV flight trajectory, the sensor wake-up scheduling and the flight time slot to ensure that the transmission information amount and energy consumption of the sensors satisfy system requirements, while maximizing the energy efficiency of the system.
Claims
exact text as granted — not AI-modifiedThe invention claimed is:
1. A design method of high energy efficiency unmanned aerial vehicle (UAV) green data acquisition system, comprising the following steps:
step 1, constructing a system optimization objective:
(1) serving a set of I ground sensors which are randomly distributed through time division multiple access (TDMA) by an unmanned aerial vehicle (UAV);
(2) flying, by the UAV, at a fixed altitude H with a maximum flight speed of V m and a total cycle of T, and discretizing the cycle T into N time slots by a time discretization method, with the length of each time slot of
t
=
T
N
;
the coordinate of the UAV is w[n]=[x(n),y(n)] T ∈R 2×1 in time slot n, wherein x(n), y(n) are the x-coordinate and y-coordinate of the UAV respectively, and R 2×1 is a two-dimensional vector space; for a sensor set SI={1, 2, . . . , I} of random distribution, the coordinate of sensor i is fixed as L i =[x i ,y i ] T ∈R 2×1 , i∈SI, each sensor supports the energy of E i ,i∈SI and the data amount to be transmitted is B i ,i∈SI; if communication between the UAV and ground is line of sight (LoS) link communication, channel quality only depends on a distance between the UAV and the sensor, and power gain in unit reference distance is expressed as ρ 0 , then the channel power gain h i [n] of the sensor i in the time slot n conforms to a free space path loss model, i.e.,
h
i
[
n
]
=
ρ
0
d
i
[
n
]
-
2
=
ρ
0
H
2
+
w
[
n
]
-
L
i
2
,
∀
i
∈
SI
,
d i [n] is the distance between the UAV and the sensor i in the three-dimensional space;
(3) assuming that the UAV serves only one sensor in one time slot, and defining a binary variable S i [n]∈{0,1} to represent wake-up scheduling of the sensor; S i [n]=1 indicates that the UAV establishes communication with the sensor i in the time slot n; S i [n]=0 indicates that the UAV does not establish communication with the sensor i in the time slot n; then, the information transmission rate R u i [n] between the UAV and the sensor i in the time slot n is expressed as:
R
u
i
[
n
]
=
S
i
[
n
]
log
2
(
1
+
P
A
ρ
0
(
H
2
+
w
[
n
]
-
L
i
2
)
σ
2
)
,
∀
i
∈
S
I
(
1
)
wherein σ 2 is additive white gaussian noise (AWGN) at a receiving end of the UAV, and P A is the transmission power of a ground sensor during communication; the total information amount R (bits/Hz) transmitted in one cycle (N time slots) of serving of the UAV is expressed as:
R
¯
(
{
W
}
,
{
t
}
,
{
S
}
)
=
∑
n
=
1
N
∑
i
=
1
I
R
u
i
[
n
]
t
(
2
)
for a rotary-wing UAV, when parameters are constant, the propulsion power P(V) of the UAV is mainly related to flight speed V; the propulsion power is composed of three parts: blade profile power, parasite power and induced power, expressed as:
P
(
V
)
=
P
0
(
1
+
3
V
2
Ω
2
r
2
)
︸
blade
profile
power
+
1
2
d
0
ρ
s
A
V
3
︸
parasite
power
+
P
i
(
1
+
V
4
4
v
0
4
-
V
2
2
v
0
2
)
1
/
2
︸
induced
power
(
3
)
the speed of the time slot n is approximately expressed as
v
[
n
]
=
w
[
n
+
1
]
-
w
[
n
]
t
=
Δ
Δ
n
t
,
and Δ n is defined as the flight distance of the time slot n; then the propulsion power P prop [n] of the time slot n is approximated expressed by the following formula:
P
prop
[
n
]
=
P
0
(
1
+
3
Δ
n
2
Ω
2
r
2
t
2
)
+
1
2
d
0
ρ
sA
Δ
n
3
t
3
+
P
i
(
1
+
Δ
n
4
4
v
0
4
t
4
-
Δ
n
2
2
v
0
2
t
(
4
)
in the formula, P 0 and P i are the blade profile power and the induced power respectively in a hovering state; Ω is blade angular velocity; r is rotor radius; d 0 represents fuselage drag ratio; ρ is air density; s is rotor solidity; A is rotor disc area; v 0 is mean rotor induced velocity; the above parameters are constants; the total propulsion energy E consumed by the UAV in one cycle of serving is expressed as:
E
(
{
W
}
,
{
t
}
)
=
∑
n
=
1
N
P
prop
[
n
]
t
(
5
)
according to the definition of energy efficiency, the system optimization objective is represented as:
EE
(
{
W
}
,
{
t
}
,
{
S
}
)
=
R
¯
(
{
W
}
,
{
t
}
,
{
S
}
)
E
(
{
W
}
,
{
t
}
)
=
∑
n
=
1
N
∑
i
=
1
I
R
u
i
n
[
t
]
∑
n
=
1
N
P
prop
n
[
t
]
(
6
)
step 2, constructing an optimization problem according to the energy efficiency formula in step 1, wherein an optimization objective is maximization of EE({W},{t},{S}), and constraints comprise UAV trajectory constraints, sensor wake-up scheduling constraints, sensor energy constraint and data amount constraint to construct the following optimization problem:
max
S
,
W
,
t
EE
(
{
W
}
,
{
t
}
,
{
S
}
)
(
7
a
)
s
.
t
.
w
[
1
]
=
w
[
N
]
(
7
b
)
w
[
n
+
1
]
-
w
[
n
]
2
≤
γ
H
2
,
n
=
1
,
…
N
-
1
(
7
c
)
w
[
n
+
1
]
-
w
[
n
]
≤
V
m
t
,
n
=
1
,
…
N
-
1
(
7
d
)
∑
i
=
1
I
S
i
[
n
]
=
1
(
7
e
)
S
i
[
n
]
∈
{
0
,
1
}
,
∀
i
∈
SI
,
∀
n
(
7
f
)
∑
n
=
1
N
(
R
u
i
[
n
]
t
)
≥
B
i
,
∀
i
∈
SI
(
7
g
)
∑
n
=
1
N
(
S
i
[
n
]
P
A
t
)
≥
E
i
,
∀
i
∈
SI
(
7
h
)
in the above optimization problem, formulas (7b)-7(d) are trajectory constraints, V m is the maximum speed of the UAV and the UAV returns to an initial position after flying by a cycle; formulas (7e) and (7f) are the sensor wake-up scheduling constraints; formula (7g) is the data amount constraint of the sensor and B i is the data amount to be transmitted by the sensor i; formula (7h) is the sensor energy constraint and E i is maximum energy supported by the sensor i in each cycle;
step 3, decomposing an original problem (7) into two sub-problems according to a block coordinate descent method; for the two sub-problems, approximately converting two non-convex problems into two convex optimization problems and calculating the problems by a successive convex approximation technique, as follows:
(1) optimization sub-problem of wake-up scheduling S and time slot t
fixing UAV trajectory W so that the sub-problem is the non-convex optimization problem of wake-up scheduling S and time slot t; firstly, for a binary variable S, slacking S to a continuous variable within a range [0,1]; then, introducing an auxiliary variable z[n] to satisfy
z
[
n
]
2
=
t
4
+
Δ
n
4
4
v
0
4
-
Δ
n
2
2
v
0
2
,
∀
n
,
i
.
e
.
,
t
4
=
z
[
n
]
4
+
Δ
n
2
v
0
2
z
[
n
]
2
,
∀
n
;
using z[n] to replace the third term of the propulsion power P prop [n] in formula (4) to obtain the UAV propulsion power P prop A [n] under the sub-problem; introducing an auxiliary variable R_t[i] to satisfy
R_t
[
i
]
2
=
∑
n
=
1
N
R
u
i
[
n
]
t
,
∀
i
;
after introducing the auxiliary variables, applying the successive convex approximation technique for non-convex constraints, converting hyperbolic constraints into SOCP and approximating the original non-convex sub-problem as a convex problem, expressed as:
max
S
,
W
,
t
z
[
n
]
,
R_t
[
i
]
.
∑
i
=
1
I
R_t
lb
[
i
]
∑
n
=
1
N
P
prop
A
[
n
]
t
(
8
a
)
s
.
t
.
w
[
n
+
1
]
-
w
[
n
]
≤
V
m
t
,
n
=
1
,
…
N
-
1
(
8
b
)
∑
i
=
1
I
S
i
[
n
]
≤
1
,
∀
n
(
8
c
)
0
≤
S
i
[
n
]
≤
1
,
∀
i
,
∀
n
(
8
d
)
R_t
lb
[
i
]
≥
B
i
,
∀
i
(
8
e
)
∑
n
=
1
N
(
S
i
[
n
]
P
A
)
≤
E
i
(
1
t
)
lb
,
∀
i
(
8
f
)
t
4
≤
z
(
r
)
[
n
]
4
+
4
z
(
r
)
[
n
]
3
(
z
[
n
]
-
z
(
r
)
[
n
]
)
+
Δ
n
2
v
0
2
(
z
(
r
)
[
n
]
2
+
2
z
(
r
)
[
n
]
(
z
[
n
]
-
z
(
r
)
[
n
]
)
)
,
∀
n
(
8
g
)
[
2
R_t
[
i
]
,
∑
n
=
1
N
R
u
i
[
n
]
-
t
]
†
≤
∑
n
=
1
N
R
u
i
[
n
]
+
t
,
∀
i
(
8
h
)
in the sub-problem (8), P prop A [n] is the propulsion power after the auxiliary variable z[n] is introduced, and is a convex function of t and z[n]; R_t lb [i] is the lower bound of first-order taylor expansion of the auxiliary variable R_t[i] 2 , and is a linear function of
R_t
[
i
]
;
(
1
t
)
lb
is the lower bound of first-order taylor expansion of
1
t
,
and has a linear relationship with t; the constraints of the sub-problem (8) are convex constraints; the optimization objective (8a) is a standard concave-convex fractional programming problem with concave numerator over convex denominator; because the constraint range is reduced by the successive convex approximation technique, the optimal solution of the convex problem after approximation is the lower bound of the optimal solution of an original sub-problem;
(2) optimization sub-problem of UAV trajectory W
fixing wake-up scheduling S and time slot t so that the sub-problem is a non-convex optimization problem of the UAV trajectory W; introducing an auxiliary variable y[n] to satisfy
y
[
n
]
2
=
t
4
+
Δ
n
4
4
v
0
4
-
Δ
n
2
2
v
0
2
,
∀
n
,
i
.
e
.
,
t
4
y
[
n
]
2
=
y
[
n
]
2
+
Δ
n
2
v
0
2
,
∀
n
;
using y[n] to replace the third term of the propulsion power P prop [n] in formula (4) to obtain the UAV propulsion power P prop B [n] under the sub-problem; after introducing the auxiliary variables, applying the successive convex approximation technique for non-convex constraints, and approximating the original non-convex sub-problem as a convex problem, expressed as:
max
W
,
y
[
n
]
∑
n
=
1
N
∑
i
=
1
I
R
u
i
,
lb
[
n
]
t
∑
n
=
1
N
P
prop
B
[
n
]
t
(
9
a
)
s
.
t
.
w
[
1
]
=
w
[
N
]
(
9
b
)
w
[
n
+
1
]
-
w
[
n
]
2
≤
γ
H
2
,
n
=
1
,
…
N
-
1
(
9
c
)
w
[
n
+
1
]
-
w
[
n
]
≤
V
m
t
,
n
=
1
,
…
N
-
1
(
9
d
)
∑
n
=
1
N
(
R
u
i
,
lb
[
n
]
t
)
≥
B
i
,
∀
i
(
9
e
)
t
4
y
[
n
]
2
≤
(
y
(
r
)
[
n
]
2
)
+
2
y
(
r
)
[
n
]
(
y
[
n
]
-
y
(
r
)
[
n
]
)
-
w
(
r
)
[
n
+
1
]
-
w
(
r
)
[
n
]
2
v
0
2
+
(
9
f
)
2
v
0
2
(
w
(
r
)
[
n
+
1
]
-
w
(
r
)
[
n
]
)
·
(
w
[
n
+
1
]
-
w
[
n
]
)
(
9
g
)
in the sub-problem (9), P prop B [n] is the propulsion power after the auxiliary variable y[n] is introduced, and is a convex function of w[n]; R u i,lb [n] is the lower bound of first-order taylor expansion of the information transmission rate R u i [n] on ∥w[n]−L i ∥, and is a concave function of w[n]; the solving method of the sub-problem (9) is the same as that of the sub-problem (8); the optimal solution of the convex problem after approximation is the lower bound of the optimal solution of the original sub-problem;
(3) overall iterative algorithm design
in each iteration, by solving the sub-problem (8) and the sub-problem (9), alternately optimizing wake-up scheduling S, the time slot t and the UAV trajectory W; using the solution obtained in each iteration as the input of next iteration; the termination condition for iteration is that the increase of optimization values of one iteration and the previous iteration is less than a set threshold; the details are as follows:
3.1) setting an iteration termination threshold ε, an initial trajectory w 0 and an iteration index r=0;
3.2) in the r+1 iteration, using the trajectory w r obtained from the r iteration to solve the sub-problem (8) to obtain the optimization result of the sub-problem (8) of the r+1 iteration, namely, wake-up scheduling S r+1 and time slot t r+1 ;
3.3) solving the sub-problem (9) by the given w r , S r+1 and t r+1 , to obtain the optimization result of the sub-problem (9) of the r+1 iteration, namely trajectory w r+1 ;
3.4) if the increase of an optimization target value is greater than a threshold ε, then updating the iteration index r=r+1; skipping back to step 4.2) for the next iteration; and if the increase of the target value is less than the threshold ε, terminating the iteration.Cited by (0)
No later patents cite this yet.
References (0)
No backward citations on record.